Time series modeling for death incidence in accidents on federal highways in Minas Gerais state, Brazil

Abstract

MinasGerais is the state with the highest deaths records by accidents on the roads under the responsibility of the Union. Therefore, we understand that it is of great importance to assess the death incidence in traffic accidents on federal highways in the state, because it can help in control and prevention actions. So, our study aims to use time series theory to model these death incidences, among January 2007 and December 2019. We verified the variability of the series, as well as the presence of the trend and seasonality components. We used AIC for choosing the best model between those that we've considered suitable. The best model to describe our time series' behavior was SARIMA(0.1.1)(0.1.1) model with just one intervention. The fitted model was used to make forecasts about future observations in this time series. According to this forecasting, we observed that deaths tend to stabilize over the months from 2020 to 2021.